By James Ashenhurst
For more pd oac 2information, please contact us. We will provide professional answers.
Last updated: July 8th, 2023 |
The Heck, Suzuki, and Olefin Metathesis Reactions Are Really Cool. [But They Don’t Belong In (Most) Introductory Courses]
There has been a trend in recent years towards including transition metal catalyzed reactions in the introductory organic chemistry curriculum. The reactions most common covered are palladium catalyzed coupling reactions (Suzuki and Heck reactions in particular) and olefin metathesis.
I generally think this is a bad idea for most courses. In my opinion, covering transition metal catalysis introduces an excessive number of “open loops” – new reactions, reagents, mechanisms, and concepts – that simply cannot be closed in any satisfactory manner given the limited time available. More on that at the bottom of the post.
However, the fact remains that the material is often covered, and students have to deal with it. Rather than completely ignore this topic, I’ve chosen to provide a condensed overview of three important reactions involving “organometallics” – the Suzuki reaction, the Heck reaction, and olefin metathesis.
First, let’s look at things from the 50,000 foot perspective.
Table of Contents
Take a look at this excellent chart of the Top 200 Pharmaceutical Products by US Volume. Here are three examples.
Each of these drugs represents the end result of incredibly expensive multi-year projects by major pharmaceutical companies and have generated untold billions of dollars in revenue. These are the “needles” found after searching through extremely large haystacks (10,000 molecules or more) in a process we call drug discovery.
This isn’t the place to cover drug discovery in detail (although if you’re curious, Derek Lowe’s In The Pipeline does that job admirably). I merely want to make the point here that organic chemists involved in this endeavour:
If you look at the molecules above, you’ll note that each of them contain aromatic rings and other sp2 hybridized carbons. This is a very common structural feature of drugs and drug-like molecules.
All this lead-up is merely to say that certain transition-metal catalyzed reactions are extremely useful for rapidly, predictably, and cheaply forming new bonds between sp2– hybridized carbons (e.g. aromatic rings) and thus have become an indispensable part of the modern organic chemists’ toolkit. They can be used to rapidly generate analogues of molecules for the purpose of improving their pharmacological properties.
[Just in case you need some perspective on the importance of small differences in structure, I should note that the mere presence or absence of a simple methyl group (CH3) has been known to sometimes make a 100-fold difference in the potency of a drug! ]
Let’s walk through the first and arguably most important transition metal catalyzed reaction we’ll cover: the Suzuki reaction.
The Suzuki reaction is a palladium-catalyzed reaction for forming a new C-C bond between two coupling partners:
• an alkenyl or aryl halide (Br, Cl, or I) [or sometimes triflate (OTf)]
• an alkenyl or aryl boronic acid R-B(OH)2 [or sometimes boronic ester, R–B(OR)2]
The reaction is catalyzed by a palladium compound, which comes in many varieties (see below) but for simplicity, we’ll use Pd(PPh3)4. Base (such as NaOH or KOH) is also essential to the reaction.
The product of the reaction contains a new C–C bond between the carbon attached to the halogen and the carbon attached to the boronic acid. The halogen and the boronic acid are removed.
You can think of palladium as acting as a “matchmaker” in this reaction – it brings the two partners together, but is itself not incorporated in the reaction product.
We can control where the bonds form by employing the appropriate precursors. For instance, in the second example above, note that our aryl bromide [1-bromo-4-methylbenzene] has a bromine “para” (opposite) the CH3 group on the aromatic ring. Had we used the “meta” isomer [1-bromo-3-methylbenzene] the new C-C bond would likewise have formed “meta” to the methyl group.
Pd-catalyzed cross coupling allows one to snap together complimentary pieces together like Lego blocks. For bonus points, you might imagine how you could use the Suzuki to build the crucial sp2-sp2 bond in Valsartan, for example.
For further reading and background on the Suzuki, I recommend the resources of Myers and White, and of course Mike Evans’ Organometallic Reader provides a general background for the key concepts of organometallic chemistry.
A second important palladium-catalyzed C–C bond forming reaction is actually the great-granddaddy of them all, known as the Heck reaction.
Just as in the Suzuki, one of the coupling partners is an alkenyl or aryl halide (Br, Cl, I) [or sometimes OTf]
Unlike the Suzuki, there is no boronic acid employed. [The names of cross-coupling reactions tend to get assigned based on the identity of the “organometallic” coupling partner]. In the Heck, the coupling partner is a mere alkene.
Here’s an example. Pay close attention to the bonds that form and break here:
The Heck reaction ends up forming a new C–C bond between the aryl halide and the alkene, and we break both a C-Br and C–H bond (that C-H bond breaking can be tricky to see!).
No doubt by looking at this one example you might find yourself coming up with questions, such as, “what’s up with all those reagents under the arrow such as Pd(OAc)2 , NEt3 and P(o-tol)3 ?”, and “why did the bond form specifically on THAT part of the alkene when there could be two other C-H bonds to break?”. I understand. More on that below.
A third important transition metal catalyzed reaction often covered in introductory organic chemistry is olefin metathesis or sometimes ring-closing metathesis. The transition metal catalyst generally employed here incorporates ruthenium (Ru) as the active metal rather than palladium.
Unlike coupling reactions, which involve alkenyl or aryl halides along with “organometal” coupling partners, the key functional group involved in olefin metathesis are very simple: alkenes.
Simply put, olefin metathesis stitches two alkenes together. It can be done between two individual alkenes (“intermolecular” or “cross-metathesis”) but is more commonly used to form rings from a molecule containing two alkene functional groups.
In the example below, some labelling will help to keep track of what bonds are being formed and broken. Note that we break the double bond between C-1 and C-2 as well as the double bond between C-7 and C-8, while forming a new double bond between C-2 and C-7 (labelled in red). Since there are 4 carbons between C-2 and C-7 we will end up forming a 6-membered ring. [Note: Ring closing metathesis works well for 5, 6, and 7 membered rings (as well as larger ones) but fails if we try to make strained 3- and 4-membered rings]
The reaction is catalyzed by a ruthenium compound which has come to be known as “Grubbs’ Catalyst”, after its discoverer, olefin metathesis pioneer (and Nobel Prize winner) Robert H. Grubbs. [I’ve taken a liberty with the structure of this Grubbs catalyst, showing the form of the catalyst when it’s active in the catalytic cycle here and not the commercially available “precatalyst”.]
There are also cases where the reverse reaction (ring-opening metathesis) can be employed, but it generally requires a cyclic alkene with some ring strain present (norbornene is a perfect example). [For the curious, you can read about ring-opening metathesis polymerization (ROMP) here.
One application of olefin metathesis has been to utilize unsaturated alkenes from plant sources (linseed oil, for example) and convert them into terminal alkenes via cross-metathesis with ethene, CH2=CH2. These terminal alkenes are commercially important: Elevance recently built a $40 million production facility to process plant oils via cross metathesis. The advantage with using a feedstock like linseed oil is that it is a renewable resource; until now, petrochemicals have typically been the hydrocarbon source for these products.
If the above seems like an overly brief coverage of three key transition metal catalyzed reactions, I agree. I actually don’t think these reactions belong in most courses, and here’s why.
First, by “cover”, I really mean, “should this be tested”. It’s one thing to tell the class to put the pens down and just have a look at these reactions which have revolutionized organic chemistry. It’s quite another to expect students to study, learn, and most importantly , understand these reactions such that they can reasonably answer questions about them on an exam.
Let’s cover the reasons “Pro” first, and then address the “Con”.
Pro: It’s reflective of modern organic chemistry.
As I described earlier, transition metal catalyzed coupling reactions (the Suzuki and Heck, among others) and Ru-catalyzed olefin metathesis are powerful reactions that get a tremendous amount of use in modern synthetic organic chemistry. By including these reactions in an introductory course, we can convey to students some of the incredible inventiveness and creativity of cutting edge organic chemists and also show them reactions that are more relevant than, say, the Cannizarro reaction or the Sandmeyer reaction, which rarely get any use these days.
Pro: Nobel Prizes were awarded for this work.
Along the same lines, the fact that these reactions have recently won Nobel Prizes for their developers (Suzuki, Heck, Negishi in 2010 for palladium-catalyzed cross-coupling; Grubbs, Schrock, and Chauvin in 2005 for olefin metathesis) is noteworthy and drives home their relevance.
Pro: If it replaces a reaction that gets little use today, like the Canizarro reaction, what’s the loss?
From a teaching perspective, spending a lecture or two on these reactions necessitates dropping some other course material. If it means skipping the Canizarro and Sandmeyer (as previously mentioned), or the malonic ester synthesis, Robinson annulation, the Ruff degradation, Hofmann elimination, and so on – is that really such a huge loss?
Also: It’s more fun to teach advanced stuff
Cross couplings and olefin metathesis represent one of the few additions to the core organic chemistry curriculum in the past 20 years. As unfair as it might sound, adding this material thus provides an opportunity to alleviate some instructor boredom, as well as to challenge students abilities to think through new material. Along the same lines, research-oriented professors often relish the opportunity to teach material that is closer to the heart of modern organic chemistry rather than spend time covering reactions advanced students are unlikely to ever encounter in the lab.
Looking at the other side:
Con – The mechanisms are completely disconnected from the rest of the course
As instructors, we value and teach the importance of understanding reactions and electron flow, not merely memorizing what is going on.
One of the fun things about teaching organic chemistry is showing students how concepts and skills that they learn in the first few weeks of class are still relevant at the very end of the course.
In every reaction from the simplest acid-base reaction to the end of the course in peptide synthesis, students can identify curved arrows and point out nucleophiles and electrophiles. The tools that are useful in chapter 4 on acid-base chemistry are still useful in chapter 24 on peptide synthesis.
Where do transition metal mechanisms fit into any of this? Yes, of course they can be understood, given enough time, and of course they do follow the core tenets of chemistry in a broad sense.
However the mechanisms themselves are so foreign to a new observer that they require a tremendous amount of work filling in the background if they are to be understood properly.
Try identifying the nucleophile and electrophile in the reaction below. How do the electrons flow?
There actually is an electrophile and a nucleophile here, but not in any sense that students are accustomed to seeing.
And this is just oxidative addition, to say nothing of other key organometallic mechanisms – reductive elimination, transmetallation, beta-hydride abstraction and so on. Covering this material is a huge diversion from the main goal of an introductory organic chemistry course: understanding electron flow.
Con: There isn’t enough time to cover the key concepts
Say you want to give students an understanding of how these reactions work. Where do you start?
In the case of the Suzuki, it’s one thing to say that it starts with oxidative addition of Pd into the C-X bond, followed by “transmetallation”, and then followed by reductive elimination. It’s another thing entirely to try to teach it in a course where students have had zero previous exposure to transition metal reaction mechanisms.
Contact us to discuss your requirements of Nano Platinum Powder supplier. Our experienced sales team can help you identify the options that best suit your needs.
And no, putting up a slide which describes the catalytic cycle of the Suzuki is not the same thing as teaching it.
Here’s a partial list of some other concepts that will be unfamiliar to someone trying to learn cross coupling reactions in the context of an introductory organic chemistry course:
Understanding these concepts is crucial to understanding how these reactions work – and I didn’t even get into such concepts as the different types of d-orbitals in transition metals or the key concept of pi backbonding.
Even in a graduate course devoted specifically to organometallics, these introductory concepts get at least a few lectures’ worth of time. But in a typical organic chemistry course, where there are only about 36 hours of lecture to go over the material, how much detail can possibly be covered?
Given the inevitable time constraints, sacrifices must be made. Often, the specific detail of the reagents involved is skipped.
How do you make the tradeoffs?
In some cases, extra details are skipped altogether. Here is an actual example of an exam question involving the Heck reaction.
What’s really being taught/learned here? Is this really so much better than teaching the ol’ Sandmeyer?
Con: It contradicts the dictum, “don’t memorize, learn the concepts”
The only way to get through Org 2 is to understand the key concepts/patterns, and not try to memorize every new reaction. But given the time constraints, how much “understanding” of organometallic reactions can possibly be achieved?
Preaching the importance of understanding concepts on one hand, and teaching the Suzuki and Heck reactions (or some watered down versions thereof) in a single day on the other, completely goes against this philosophy.
There’s better ways to include more advanced material without introducing all the new concepts that come with transition metal chemistry.
Some examples that come to mind are introducing concepts such as hyperconjugation, orbital symmetry, steric and stereoelectronic effects. These can augment and inform students’ understanding of material they are already learning without burdening them with a overwhelming number of “open loops” that transition metal chemistry introduces.
Next post: Reaction Map – Reactions of Organometallics
These reactions, which are based on fundamental organometallic and inorganic chemistry principles, are indeed challenging to teach in a traditional undergraduate introductory organic chemistry course. That being said, they are commonly included in most curricula nowadays and they are more or less here to stay.
For more information on each of these reactions, here are some links to the relevant entries in the Reaction Guide:
Here are some more pedagogical papers on the above topics:
We chose to investigate the reaction of 2-bromo-N-phenyl benzamide 1a, primarily because of the potential amide-directed C-H activation of the N-phenyl moiety. Given the ability for a range of Pd pre-catalysts and monodentate ligands to catalyze the formation of phenanthridinones 2 (Fig. 1), we planned to assess the ability of bidentate ligands to affect the transformation, particularly as their characterization in follow-on mechanistic studies (i.e., stoichiometric experiments with Pd) might be more facile. An initial pre-catalyst and ligand screen (see Supplementary Information for full details) showed that 5 mol% Pd(OAc)2 {formally high purity Pd3(OAc)6, nitrite-free}38 and dppe, with K2CO3 as base in DMF39 at 80 °C gave N-phenyl phenanthridinone 2a in 58% yield (Fig. 2C). Furuta / Kan17 and Fu / Chen22 obtained 2a in 23% and 80% yield respectively from 1a. We also determined that dppp could be used in place of dppe giving similar results. Crucially, for application to HTE, this reaction was found to be tolerant of air when employing the Pd(OAc)2/dppe pre-catalyst. Furthermore, conducting the reaction under anhydrous and oxygen-free conditions gave a yield of 2a which was 54%.
Since the right-hand disubstituted aryl ring in 2a could be incorporated from either aryl group of 1a we sought to confirm that it originates from the bromobenzene moiety (rather than the N-phenyl ring). As expected, use of deuterated benzamide 1a-d5 gave phenanthridinone 2a-d5 as a single product in 61% yield after 2 h at 130 °C. We further probed the regioselectivity of the reaction by employing methyl-substituted benzamide 1b, which afforded phenanthridinone 2b as a single regioisomeric product in 39% yield, the structure of which was confirmed by single crystal X-ray diffraction analysis and NMR spectroscopic analysis. This molecular substitution pattern confirms20 the mechanism proceeds through a biaryl intermediate to D (Fig. 2A), under our catalytic reaction conditions.
The reaction of 1a mediated by Pd(OAc)2/dppe (1:1, 5 mol%) was monitored in operando using IR spectroscopic analysis, with the consumption of 1a assessed at 80 °C in DMF using both a cold and hot pre-catalyst activation protocol (Fig. 2D). When the reaction mixture was heated from room temperature (ca. 20 °C), an induction period of ca. 10 min was evident from the color changes recorded (yellow→orange→dark red; the latter indicative of Pd0 species forming, vide infra), with the reaction nearing completion within 2.25 h. The analysis is complicated by the complete dissolution of 1a requiring heating (which can be seen by the appearance and then subsequent disappearance of 1a). Heating the pre-catalyst mixture in DMF to 80 °C for 2 min prior to addition of 1a shortens the induction period considerably (<2 min; reaction mixture a dark red color), with the reaction reaching completion within <2 h (confirmed by independent 1H NMR analysis). The difference in the pre-catalyst activation process highlights the impact on the overall reaction time. Independent reactions between Pd(OAc)2 and dppe showed that Pd0(dppe)n species (n = 1 or 2) were formed (vide infra).
Next, a robust LC-MS method for profiling the reactions from the HTE campaign (in batch mode) was developed. Generally, we observed (by chromatographic and spectroscopic methods) that many products were formed in reaction mixtures accompanying 2a, 2a-d5, and 2b, the majority of which were formed in low amounts but in significant enough quantity to warrant comprehensive profiling. This would fulfill our aspirations to gain greater understanding of this complex reaction network. Characterization of these species was achieved by LC-MS, GC-MS, flash column chromatography and preparative HPLC on the crude reaction mixtures and comparing the spectroscopic and chromatographic data with those of authentic product samples (see Supplementary Information for full details). Through this approach, we identified a total of 17 side-products and by-products of interest in the LC-MS reaction profiles (Fig. 3), many more than were revealed than in the several previous studies16,17,18,19,20,21,22.
Fig. 3: A full survey of the reaction products detected by analytical methods (LC-MS and GC-MS).The blue color in compounds shows the origin of the benzo-moiety.
Full size image
Careful column chromatography resulted in the isolation and characterization of major side-products including symmetrical biaryl 4a, isolated in ~10% yield (analogous with 4, Fig. 2B), amide 6a16, pentacycle 7a40,41, and phenol 8a. Amides 5a and 6a are analogous to compounds 5 and 6 (Fig. 2B) identified by ref. 20 and ref. 16 respectively. Porée et al. showed that phenanthridinones 5 (where R’ = benzyl, methyl) were formed in presence of certain base cations and solvents19,20, for which definitive conclusions could not be drawn. We examined the reaction of 1a using Porée’s pre-catalyst PdCl2(PPh3)2 (5 mol%) with K2CO3 (3 equiv.) in dioxane at 105 °C, however, 5a was only formed in a trace quantity (by HPLC) in our hands.
Substituted phenol 8a possibly originates from reductive elimination from a “Ar-Pd-OH” species42, facilitated by the presence of residual water in the reaction medium. Interestingly, the formation of 8a indicates the presence of hydroxide in the system, akin to the Suzuki-Miyaura cross-coupling43,44. N-phenyl benzamide 9a likely occurs due to proto-dehalogenation, a side reaction that is common in Pd-catalyzed reactions in basic DMF reaction media, but not fully understood mechanistically45. Biaryl amide 10a, a formal cross-coupled product, was also identified. Interestingly, experiments employing deuterated 1a-d5 and methylated benzamide 1b starting materials indicated that the “new” aryl ring does not originate from 2-bromo-benzamide 1, but instead originates from the phosphine ligand (see Supplementary Information for full details). Reactions of phosphine ligands (e.g., PPh346 and dppe47) with substrates and products at Pd are known, either via liberation of phenyl moieties and phosphido (–PPh2) or phosphonium (+PPh4) ions2,3.
Other side-products identified include 11a, presumably formed through a Buchwald-Hartwig amination type reaction of 2-bromo-N-phenyl benzamide 1a with aniline 16a; arylated product 12a, which presumably arises through amide-directed C-H arylation of phenanthridinone 2a (as with biaryl 10a, the “new” aryl ring originates from the dppe ligand); and 13a, a hydrolysis product of 4a. Side-products 14a and 15a were observed when methyl ethyl ketone (MEK) was used as a solvent and likely arise through the mono- and di-α-arylation of MEK respectively48,49,50.
Furthermore, symmetrical urea 3a was observed by LC-MS along with the requisite building block, aniline 16a produced during ipso-substitution (D to E, Fig. 2A)16,17. Another aniline derived side-product, acetanilide 17a, was identified in the analytical LC-MS method. We believe that this is formed by a Pd-mediated acetylation processs5. Quinazolinedione 18a51 could be formed through a reaction of the substrate 1a with phenyl isocyanate 20a (vide infra) and then a subsequent intramolecular amination cyclization process, or a carbonylative-type process.
We observed 19a by LC-MS, which was present throughout the HTE campaign (m/z = 469). However, we were unable to delineate its structure (see Supplementary Information for full details). Thus, we have treated 19a as being an unknown species. As we demonstrate below, our approach to complex reaction analysis allows the effects of changing reaction parameters on this unknown species / contaminant to be uncovered.
Finally, we identified several species by GC-MS of the reaction mixtures that were invisible to the LC-MS method (Fig. 3). These include the expected by-product phenyl isocyanate 20a, azobenzene 21a from oxidative coupling of aniline, which is likely promoted by Pd nanoparticles52, hydrolysis product 22a and bi- and terphenyls 23, 24, and 25. Whilst these species were not included in the profiling of reaction from our HTE work vide infra, their presence confirms the high degree of complexity of the reaction under study. Moreover, the identification of these species via GC-MS, but not LC-MS highlights that the sole use of one analysis method (which is commonplace in HTE reaction screening) ought to be viewed with some caution.
With a useful LC-MS method established, we next explored the effect of changing reaction conditions on the reaction profile using HTE. Given that most reported examples of phenanthridinone 2 syntheses use polar aprotic solvents (typically DMF and DMA) at temperatures between 100–150 °C (Fig. 1), alongside the solvent effects observed by refs. 19,20, we selected to study the effect of solvent and temperature in further detail. To this end we selected seven separate solvents {DMF, propylene carbonate, acetonitrile, n-butyronitrile, methyl ethyl ketone (MEK), di-n-butyl ether and toluene} that covered a wide range of polarities, as well as a 9:1 n-Bu2O:DMF mixture (the latter solvent mixture to aid greater solubility of the reaction components). Furthermore, propylene carbonate53,54, and MEK55 were selected primarily as potentially greener and less toxic alternatives. Five temperatures between 80–120 °C (10 °C intervals) were chosen, as were several reaction time-points from 1 to 8 h to provide a temporal visualization of the altering product(s) profile. Reactions were conducted at 130 °C, but significant solvent losses were noted in these experiments leading to unreliable data. We elected to perform the HTE campaign without a catalyst pre-activation step (see Fig. 2D) to avoid variability in the experimental workflow, which facilitated the range of side-products generated to be fully explored.
Reactions were performed on a Chemspeed ISYNTH robotic platform with a solid-dispensing unit14,15,56,57,58 to expedite reaction set-up and sampling and off-line LC-MS analysis was used to generate the reaction profile (see Supplementary Information for full experimental details, including workflow schematic diagram). We rigorously assessed positional variability and reproducibility using the ISYNTH system. In total, 40 reactions were performed in triplicate, with sampling at four different time points, generating 480 reaction profiles. LC-MS alignment and peak picking were performed using automated processing (Progenesis QI)59 and relative concentrations of the reaction constituents were obtained by normalizing the mass-ion counts over the peaks of interest (see Supplementary Information for full details), resulting in a semi-quantification of the species of interest. This approach allows the study of the variation in amount of these (characterized and uncharacterized) species with changing reaction conditions and is simpler than developing an analytical method to show absolute quantities of all species (which would be a non-viable effort in most academic/industrial settings for reactions of this complexity).
Analysis of such large multidimensional datasets is difficult without multi-variate statistical analysis techniques to reduce dimensionality. To visualize the effects of reaction variables, we employed principal component analysis (PCA)60, an unsupervised data analysis method that allows patterns in data with multiple variables to be observed in scatter plots with minimal loss of information. This is achieved by a rotation of the multi-dimensional axes, where each axis corresponds to a different variable, in our case the integrated intensities of identified peaks from LC-MS analysis. The rotation preserves the orthogonality of the axes but the first new axis, or principal component, is chosen as the direction of maximum variance in the data, the second principal component corresponds to the next most variance (orthogonal to the first), and so on. In this way, a small subset of the new axes, or principal components, provides most of the information in the data and pairwise scatter plots showing the coordinates of the observations in relation to even just the first two principal components can reveal meaningful patterns in the data. As these new axes are obtained by a rotation, each principal component is a linear combination of the original variables or peak intensities, and the rotation matrix provides the coefficients, or loadings, of each in the linear combinations. The magnitudes of the loadings for a particular component therefore show the importance of each species to any patterns observed in the direction of that component.
We first examined data obtained at a single temperature to reduce complexity and allow the effect of time and solvent to be studied independently of temperature. At 110 °C these effects are pronounced and highlight important trends. Indeed, PCA shows clear differences between solvents (Fig. 4). For PCA plots including all times and temperatures see the Supplementary Information. The insets show the loadings for the first two principal components (PC1 and PC2) as vectors, describing the contribution of the different compounds to the patterns seen in the corresponding scores plots.
Fig. 4: Scores plots for the first two principal components obtained with UV-scaled data from experiments performed in 8 different solvent systems.The plot in (A) shows experiments performed at 110 °C for four different reaction times whereas (B) shows 2-h reactions performed at five different temperatures. The loadings, shown as vectors in the insets, indicate the contribution of the various products to the principal components.
Full size image
As the variance for large peaks is greater than that for small peaks, the major products dominate the analysis, and differences due to small peaks are masked unless the variables are rescaled. UV-scaling, or scaling to unit variance gives all variables equal influence on the analysis. PCA scores plots for unscaled data are shown in the Supplementary Information. Here, differences along PC1, accounting for 97.3% of the variance in the data, are related to differences between 2-bromo-benzamide 1a, the starting material, and major product 2a. As the variance involves squared values, the scores plot can be mirrored along either PC1 or PC2 and needs to be oriented by looking at the data used in the analysis (provided as supplementary material). In this case, the data shows that Bu2O observations have the highest values of 2-bromo-benzamide 1a remaining and MeCN observations have the highest values of product 4a. Thus, positive PC1 scores are associated with high amounts of 2a and negative scores with high amounts of 1a, whilst positive PC2 scores are associated with the higher levels of other products.
After scaling the data so the analysis is not dominated by the large amounts of 2-bromo-benzamide 1a and phenanthridinone 2a, PCA shows the distribution of side-products associated with different solvents (Fig. 4). After orienting the scores plot using the original data, it can be seen that, while DMF and propylene carbonate gave the greatest amount of 2a, regardless of reaction time, Bu2O led to the most 1a remaining, with a slow but steady increase in product 2a with increasing reaction time (Fig. 4A). Other solvents, notably toluene, show a dependence on reaction time with most 1-h reactions in the center of the plot. Bu2O and toluene have the lowest values of most products, except perhaps products 7a, 8a, 9a and 12a. The large positive loadings along PC1 for the major product 2a, as well as aniline 16a, and amides 5a and 6a show that DMF and propylene carbonate observations on the right of the plot are associated with greater quantities of these products. In fact, further analysis (see Fig. 5) shows that, while both solvents produce high amounts of the major product 2a, propylene carbonate produces higher quantities of 6a whereas DMF is more associated, with higher quantities of 5a and 16a.
Fig. 5: Scores plots for the first two principal components obtained after removing outliers from UV-scaled data from experiments performed in 8 different solvent systems.The plot in (A) shows experiments performed at 110 °C for four different reaction times whereas (B) shows 2-h reactions performed at five different temperatures. In both cases, 13 observations were removed. The loadings, shown as vectors in the insets, indicate the contribution of the various products to the principal components.
Full size image
When considering only 2-h reaction time points performed at different temperatures, PCA reveals similar patterns with solvent (Fig. 4B). Again, reactions using Bu2O have most 1a remaining. Except for the lowest temperature reactions (reactions at 80 °C in DMF and 110 °C in 9:1 n-Bu2O/DMF), DMF and propylene carbonate result in the most phenanthridinone 2a. Most solvents show higher conversion of 1a with an increase in temperature. The greater quantities of side-products in MeCN and MEK can be seen again but outcome for both BuCN and the dual solvent system is variable. In addition to product 2a, higher quantities of aniline 16a and 6a could be associated with DMF and propylene carbonate, whilst Bu2O and toluene have the lowest values. Much of the variance along PC2 is due to unusually high proportions of some compounds being recorded for a few observations, emphasized by the scaling. As the replicate experiments did not have similar outcomes, these unusual observations could be considered outliers due to reaction sampling or analysis errors.
To show any trends with reaction time or temperature more clearly, outliers were removed, and the analysis repeated. Outliers were determined by considering the similarity of replicates. For each set of replicates, the Euclidean distance to the centroid was calculated. The mean distance plus 1.5 standard deviations was set as a threshold and any replicates with a distance greater than this were removed. This threshold was chosen as it removed the worst outliers from each dataset without taking out too many observations. Figure 5 shows the results for scaled data. The reaction time data now shows that MEK is most associated with greater amounts of a combination of products 10a, 11a, 14a, 15a and 3a, while MeCN, BuCN and the dual solvent system are more related with greater amounts of products 18a and 19a (Fig. 5A). The temperature data set shows an even greater difference in the amount of these side-products for MEK and the increasing trend with temperature can be seen more clearly for MeCN, BuCN and the dual solvent system.
Although PCA gives a good indication of the products associated with the different solvents, interpretation is complicated when several variables have similar loadings. Importantly, it is the sum of the variables that contributes to the principal component in question. For example, a lower value for side-product 6a may be compensated for by a higher value for aniline 16a, leading to a similar score as a higher value for 6a with a lower value for 16a (Fig. 4B). To determine more specific relationships between solvents and reaction products, we employed correspondence analysis (CA)33. The median quantities were calculated over replicate reactions and, for each solvent, the number of observations counted with high, medium, and low quantities of each product, overall temperatures and reaction times (a total of 20 for each solvent). Quantities less than a third of the full range (i.e., the maximum minus minimum value over all solvents) for a product were defined as low, quantities between a third and two-thirds as medium, with quantities above two-thirds of the range for the product considered as high (see Table 1).
Table 1 The range of values for each compound. Values below the low/medium threshold were considered “low” and values above the medium/high threshold were considered “high” whilst values between the two thresholds were considered “medium”Full size table
The results are shown in Fig. 6A. The strength of the association between solvents and reaction products depends on the distance from the origin (where the dotted lines cross) and on the angle between the vectors from the origin to the points representing the solvents and products. As in PCA, the association of DMF and propylene carbonate with high amounts of the phenanthridinone product 2a can be seen, but here acetanilide 17a is associated with propylene carbonate, whilst 5a and aniline 16a are associated with DMF. Furthermore, 6a is associated with both propylene carbonate and, to a lesser extent, DMF.
Fig. 6: Further data analysis.A Correspondence analysis74 biplot showing associations between solvents and reaction products. Key: Dim = dimension. The number of experiments (after combining replicate analyses) with high (black), medium (orange), and low (light green) quantities for each product are used in the analysis. B Bubble plot showing the number of experiments with quantities above two-thirds of the range for the product by solvent. Bubble sizes are proportional to the number of experiments, also shown where the maximum possible is 20 (i.e., 5 temperatures for 4 reaction times).
Full size image
The biplot also shows that MeCN is related to relatively high amounts of Buchwald-Hartwig product 11a, whilst MEK is most associated with MEK α-arylation products 14a and 15a (as expected for the latter solvent which serves as a substrate). Other solvents, being closer to the origin, are less well discriminated.
The bubble plot in Fig. 6B provides another means for data visualization. Here, only the number of experiments with high quantities are included. It can be readily seen that product 5a is associated with reaction in DMF with 16 of the 20 reaction profiles recorded when DMF was used (5 temperatures for 4 reaction times) show relatively high quantities of 5a. Conversely, when the reaction is performed in propylene carbonate, no reactions exhibit high quantities of 5a. However, 14 of the 20 reactions show high levels of 6a. Additionally, 12 reactions show relatively high amounts of acetanilide 17a. Again, that high amounts of α-arylation products 14a and 15a are only associated with MEK, which is as expected, validating our approach.
To garner further mechanistic insight, we used heatmaps to visualize correlations between products. The heatmap in Fig. 7 shows the strength of any correlation across all reactions between the various products. Therefore, the correlation of a product with itself along the diagonal has the strongest positive correlation. The data includes reaction solvents, temperatures, and times. For example, the similarity of MEK α-arylation products 14a and 15a is clear from the dark red block indicating a positive correlation close to 1 (highlighted by a yellow box). The strongest negative correlations (dark blue) are between 2-bromo-benzamide 1a, and the block of positively correlated products, aniline 16a, biaryl diamide 4a, 2a-derivatized compound 6a, and phenanthridinone 2a (the major product of the reaction). This negative correlation highlights that significant consumption of 1a is correlated with formation of phenanthridinone 2a and the most prevalent side products 4a and 6a.
Fig. 7: Heatmap showing correlations between products across reactions, including all solvents, reaction times and temperatures.Median values of replicate observations were used in the analysis. Key: red boxes are positive correlations and blue boxes are negative correlations (see color bar, upper left). The reaction products are ordered using hierarchical clustering, resulting in the dendrograms shown in the margins, so that similar products cluster together. The map colors show the strength of correlation, as indicated by the color bar. Specific clusters of compounds are highlighted by colored boxes, representing selected interactions.
Full size image
Aniline 16a correlates with carboxylic acid 13a (with both possibly formed from the hydrolysis of 4a) and 5a. Oxidative reaction of 2a (by C-H bond activation) with urea 3a could in principle lead to both 5a and 16a, leading to positive correlations.
Other blocks of positively correlated products show 7a (oxidative cyclization product) and 9a (reductive proto-debrominated product) grouped together. Compound 11a (the Buchwald-Hartwig amination product) correlates with both 10a (phenylated product) and quinazolinedione 18a. Although not highlighted, Fig. 7 also shows the positive correlation of 5a and 16a with 2a, supported by the similar loadings in PCA and their proximity in correspondence analysis (mainly due to being produced together in DMF reactions). Such data correlations form the basis for mechanistic proposals (vide infra).
To supplement our mechanistic predictions about the catalytic cycle(s) operative in this complicated chemistry, we recognized that our HTE/data analysis would be complemented by traditional stoichiometric studies involving appropriate Pd precursor compounds. The aim of this part of the study was not to provide definitive mechanistic information, but to support the connections made in the HTE/data analysis. Previously proposed catalytic cycles reported by ref. 20. were based primarily on DFT calculations using PH3 as a model for the PPh3 ligand (along with an N-methyl substrate variant of 1a). Such a model makes a PdII/PdIV catalytic cycle possible, in principle (Fig. 2A). While higher oxidation state PdIV intermediates are experimentally feasible24, evidence for the stabilization of PdIV species by phosphines is relatively limited, exceptions being PPh361 and transphos62. Indeed, the scientific community at large have often doubted Pd catalytic cycles involved phosphine-stabilized PdIV intermediates63,64. While dppe could be related to these ligands, we conducted stoichiometric 31P NMR and MS experiments to better understand dppe interactions at Pd in the catalytic system of interest 1a → 2a and other side-products and by-products.
Amatore and Jutand showed that Pd0(dppe)2 I’ and Pd0(η2-dba)(dppe) I-dba are formed from the reaction of Pd2(dba)3•dba with n dppe (n = 1 or 2) in THF65. In our hands, reaction of 1 equivalent of Pd(OAc)2 with 1 equivalent of dppe in DMF-d7 at 23 °C resulted in a mixture of Pd0(dppe)2 I’ (δ 30.10 ppm; lit. δ 30.46 ppm in THF) and PdII(dppe)(OAc)2 (δ 59.07 ppm; lit66. δ 58.9 ppm in CH2Cl2) being formed (Fig. 8). There was no evidence for phosphine oxidation or P-C bond cleavage (forming potential phosphido groups) in these experiments. As ligand dissociation from Pd0(dppe)2 I’ is thought to generate the putative unsaturated (catalytic) Pd0 species Pd0(dppe) I, evidence was gained by addition of dba to the mixture of PdII(OAc)2/1 dppe, and cross-referencing to an authentic sample of Pd0(η2-dba)(dppe) I-dba (δ 35.5 and 37.0 ppm, ∆ν1/2 = 28 Hz)65. We also note that the presence of dba in the catalytic reaction does not significantly perturb the reaction system in terms of the many products that are formed (i.e., 0.5Pd2(dba)3/1dppe is a component catalyst system). The experiment confirmed Pd0(dppe) liberation from Pd0(dppe)2 I’ in DMF-d7 at 23 °C; this latter species is an observable species, acting as a catalyst reservoir.
Fig. 8: Stoichiometric palladium reactivity studies using substrate 1a.31P NMR (203 MHz, DMF-d7) spectral changes and confirmation of derived (pseudo)molecular ions by ESI-MS: (1) Pd2dba3·CHCl3 and dppe taken at rt, t = 0 min; (2) taken after addition of 2-bromo-benzamide 1a at rt, t = 5 min; (3) taken after 45 min heating at 80 °C; (4) taken after 16 h heating at 80 °C. We have assigned species to five phosphorus species, cross-referenced with the ESI-MS analysis (simulated ions shown in red; experimental ions in blue). The species at δ 57.35 is a trace product, derived from the formation of Pd0(dba)(dppe)/Pd0(dppe)2, which is distinct to the species at δ 56.77 ppm, formed at higher temperature, assigned to PdBr2(dppe) VIII.
Full size image
We assessed oxidative addition of 2-bromo-N-phenylbenzamide 1a to a pre-synthesized mixture of Pd0(dppe)2 I’ (Fig. 6, red) and Pd0(η2-dba)(dppe) I-dba (Fig. 8, orange), derived from Pd02(dba)3•dba and 1 dppe in DMF-d7 at 23 °C. It was necessary to heat the mixture of 1a with these Pd0 species at 80 °C to give key oxidative intermediate II (Fig. 6, yellow). Two doublets appeared at δ 41.78 and δ 59.05 ppm (2JPP = 28 Hz), supporting formation of PdII(Ar)Br(dppe) II. The 31P NMR signals for oxidative addition products of this type are typically found in the δ 56–31 ppm range (in THF or DMF)67,68. In our case, the molecular fragment [M-Br]+ was verified by ESI-MS (+ve mode). A new singlet peak at δ 65.83 ppm (Fig. 8, blue) was observed, which we attribute to reaction intermediate VI, resulting from the formation of a C-C bond between two benzo-groups. This proposal is supported by the detection of the sodiated pseudo-molecular ion [VI + Na]+ by ESI-MS (+ve mode).
Another new species with a signal of δ 56.55 ppm (Fig. 8, green) forms at 80 °C, which we tentatively assign to PdII(dppe)(Br)2 VIII (Fig. 6, green). In a separate reported study an autoionization equilibrium was noted for the behavior of PdII(dppe)(OAc)2 to give [PdII(dppe)2]2+ VIII’ + Pd(OAc)266. The detection of [PdII(dppe)2]2+ VIII’ by ESI-MS (+ve mode) allows us to assign this signal to PdII(dppe)(Br)2, VIII with a similar autoionization equilibrium [PdII(dppe)2]2+ VIII’ and PdBr2 in play66. The formation of Pd black was noted at the end of this stoichiometric reaction, which is typical for these types of experiments.
Based on the complete information to hand we propose mechanistic hypotheses, as outlined in Fig. 9A. Firstly, for the catalyst activation step we have demonstrated that Pd0(dppe) I species is generated under the reaction conditions, as the likely Pd0 species in solution (in DMF), for which there are other potential reductants in the system (thus not dependent on the phosphine ligand per se, Fig. 9A). The generation of aniline 16a under the working catalytic reaction conditions could assist formation of Pd0, leading to the generation of dppe mono-oxide69 (an alternative ligand, vide infra) and acetanilide 17a, generated through acetate transfer to aniline 16a. Given the higher reaction temperatures used for the catalytic reaction there are different mechanisms possible for pre-catalyst activation and formation of Pd0 species (see Supplementary Information, Figure 15). We expect in polar aprotic solvents, e.g., DMF, that the presence of trace dimethylamine, CO and water would likely influence the reductive process.
Fig. 9: Detailed catalytic cycle for the formation of phenanthridinone 2a.Key: O.A. oxidative addition, R.E. reductive elimination, RDS is rate determining step. We expect all steps resulting in loss of HBr to involve base. Potential catalytic cycles for selected side-products, based on the correlations revealed by the rich data analysis (from heat maps and hierarchical clustering; highlighted by appropriate colors). We expect all steps resulting in loss of HBr to involve base. We do not preclude dppeO being an alternative ligand for Pd in these catalytic cycles, especially in solvents where there is no other potential reductant, i.e., where phosphine ligand becomes the obvious reductant. The blue color in the chemical structures shows the origin of the benzo-moiety.
Full size image
We have confirmed experimentally that oxidative addition of 2-bromo-N-phenylbenzamide 1a with Pd0 species containing dppe occurs to give PdII species II containing new Pd-Br and Pd-C bonds (confirming scission of the C-Br bond in 1a). From this point, a second oxidative addition-type reaction (mechanism not characterized) of 2-bromo-N-phenylbenzamide 1a to II is necessary to generate a putative PdIV species III (following the mechanism20 hypothesized by a Porée et al.). This is critical in generating a new C-C bond between the required benzo-groups, which occurs by subsequent reductive elimination to give PdII intermediate IV. The structure of VI is supported by NMR and MS, thus IV is connected by loss of two equivalents of HBr via V (which is likely base-assisted). This could in principle occur in a single step. We propose that VI is involved in the rate determining step/state (RDS). Extrusion of phenylisocyanate 20a (which can be trapped by reaction with aniline 16a to afford the urea 3a) then affords PdII intermediate VII, which we believe to possess a high energetic barrier. An alternative pathway from V → VII is shown (via V´), involving amide hydrolysis, formation of a Pd-O bond, extrusion of CO2 and formation of aniline 16a. The final step involves reductive elimination from VII to give phenanthridinone 2a (the major and dominant product), with concomitant regeneration of the active Pd0 catalyst species “Pd0(dppe)” I.
Figure 9 further reveals how other major side-products 4a (main cycle, de-coordination from PdII intermediate IV), 5a, 6a, and 7a are connected to the main catalytic cycle. The heat map and hierarchical cluster analysis connecting 5a, 13a, and 16a, which are formed as downstream products from proposed intermediate IV, indicate that there are several exit points from the main catalytic cycle from which these side-products form.
We propose that pentacycle 7a arises through a formal oxidative C-H activation of phenanthridinone 2a. Thus, reaction of Pd(dppe)Br2 VIII with 2a would involve formation of PdII intermediate IX (loss of HBr). Cyclopalladation, with loss of HBr70 would give 6-membered ring palladacycle X. A classical reductive elimination step then forms 7a and leads to the regeneration of ‘Pd0(dppe)’ catalyst species I. The confirmed appearance of Pd(dppe)Br2 VIII in our stoichiometric Pd experiments supports this potential oxidative process. The association of proto-debrominated product 9a with the formation of 7a (revealed by the Heat Map and hierarchical cluster analysis) suggests that 2-bromo-N-phenylbenzamide 1a is capable of acting as an oxidant in this competing catalytic cycle. Organohalides acting as oxidants for Pd0 → PdII is established and has been exploited in other chemical synthesis applications71,72. We do not preclude a role for trace air in this process, serving a role to oxidize “Pd0” to “PdII” species that would be equally capable of promoting the formation of 7a through a similar sequence of steps. Qualitatively we can state that in the presence of air 7a is formed in higher quantities.
The heat map correlation of side-product 6a with 2a allows us to propose that a C-H activation process involving phenanthridinone 2a and PdII oxidative addition intermediate II is likely, giving complex XI (detailed steps not given—cycle shown in Fig. 9). Subsequent reductive elimination regenerates the “Pd0(dppe)” catalyst species I, releasing compound 6a as another side-product. Catellani et al. has suggested that compounds similar to 6 could arise by intramolecular cyclopalladation, followed by coupling with 1a and ipso-substitution16. We recognize that 6a could derive directly from intermediate VI, through phenanthridinone ring-formation, retaining the phenyl amide group. This would account for 6a having positive interactions with 2a in the correlation.
Building on the discussion above about the correlation of 5a, 13a, and 16a—Porée et al. found20 that compounds like 5 dominated under reaction conditions in which bromine–carbonate exchange and hence, rotation around the biaryl axis is disfavored. We acknowledge that 5a could be formed by reaction of urea 3a with product 2a, for which the heat map analysis shows a strong positive correlation. Hydrolysis of 4a likely occurs at PdII, thus carboxylic acid 13a and aniline 16a derived from intermediate IV.
The dppe ligand is a phenyl donor source that is transferred to 10a and 12a—it is established that phenyl-containing phosphines are capable of transferring a phenyl group at Pd47. On the other hand, hydrolysis product 8a likely derives from oxidative addition intermediate II. The heat map correlation between 10a, 11a, and 18a shows that phenyl transfer from the dppe ligand is prevalent where amination occurs. Quinazolinedione 18a formation is linked to 11a, which we propose to be formed via a carbonylative process.
Are you interested in learning more about oled materials? Contact us today to secure an expert consultation!